def process_request(self, request):
"""Process the given request"""
asset_path = request.path
if self.is_asset_request(request):
# Make sure we can convert this request into a location.
if AssetLocator.CANONICAL_NAMESPACE in asset_path:
asset_path = asset_path.replace('block/', 'block@', 1)
# If this is a versioned request, pull out the digest and chop off the prefix.
requested_digest = None
if StaticContent.is_versioned_asset_path(asset_path):
requested_digest, asset_path = StaticContent.parse_versioned_asset_path(asset_path)
# Make sure we have a valid location value for this asset.
try:
loc = StaticContent.get_location_from_path(asset_path)
except (InvalidLocationError, InvalidKeyError):
return HttpResponseBadRequest()
# Attempt to load the asset to make sure it exists, and grab the asset digest
# if we're able to load it.
actual_digest = None
try:
content = self.load_asset_from_location(loc)
actual_digest = content.content_digest
except (ItemNotFoundError, NotFoundError):
return HttpResponseNotFound()
# If this was a versioned asset, and the digest doesn't match, redirect
# them to the actual version.
if requested_digest is not None and (actual_digest != requested_digest):
actual_asset_path = StaticContent.add_version_to_asset_path(asset_path, actual_digest)
return HttpResponsePermanentRedirect(actual_asset_path)
# Set the basics for this request. Make sure that the course key for this
# asset has a run, which old-style courses do not. Otherwise, this will
# explode when the key is serialized to be sent to NR.
safe_course_key = loc.course_key
if safe_course_key.run is None:
safe_course_key = safe_course_key.replace(run='only')
newrelic.agent.add_custom_parameter('course_id', safe_course_key)
newrelic.agent.add_custom_parameter('org', loc.org)
newrelic.agent.add_custom_parameter('contentserver.path', loc.path)
# Figure out if this is a CDN using us as the origin.
is_from_cdn = StaticContentServer.is_cdn_request(request)
newrelic.agent.add_custom_parameter('contentserver.from_cdn', is_from_cdn)
# Check if this content is locked or not.
locked = self.is_content_locked(content)
newrelic.agent.add_custom_parameter('contentserver.locked', locked)
# Check that user has access to the content.
if not self.is_user_authorized(request, content, loc):
return HttpResponseForbidden('Unauthorized')
# Figure out if the client sent us a conditional request, and let them know
# if this asset has changed since then.
last_modified_at_str = content.last_modified_at.strftime(HTTP_DATE_FORMAT)
if 'HTTP_IF_MODIFIED_SINCE' in request.META:
if_modified_since = request.META['HTTP_IF_MODIFIED_SINCE']
if if_modified_since == last_modified_at_str:
return HttpResponseNotModified()
# *** File streaming within a byte range ***
# If a Range is provided, parse Range attribute of the request
# Add Content-Range in the response if Range is structurally correct
# Request -> Range attribute structure: "Range: bytes=first-[last]"
# Response -> Content-Range attribute structure: "Content-Range: bytes first-last/totalLength"
# http://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.35
response = None
if request.META.get('HTTP_RANGE'):
# If we have a StaticContent, get a StaticContentStream. Can't manipulate the bytes otherwise.
if type(content) == StaticContent:
content = AssetManager.find(loc, as_stream=True)
header_value = request.META['HTTP_RANGE']
try:
unit, ranges = parse_range_header(header_value, content.length)
except ValueError as exception:
# If the header field is syntactically invalid it should be ignored.
log.exception(
u"%s in Range header: %s for content: %s", exception.message, header_value, unicode(loc)
)
else:
if unit != 'bytes':
# Only accept ranges in bytes
log.warning(u"Unknown unit in Range header: %s for content: %s", header_value, unicode(loc))
elif len(ranges) > 1:
# According to Http/1.1 spec content for multiple ranges should be sent as a multipart message.
# http://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.16
# But we send back the full content.
log.warning(
u"More than 1 ranges in Range header: %s for content: %s", header_value, unicode(loc)
)
else:
first, last = ranges[0]
if 0 <= first <= last < content.length:
# If the byte range is satisfiable
response = HttpResponse(content.stream_data_in_range(first, last))
response['Content-Range'] = 'bytes {first}-{last}/{length}'.format(
first=first, last=last, length=content.length
)
response['Content-Length'] = str(last - first + 1)
response.status_code = 206 # Partial Content
newrelic.agent.add_custom_parameter('contentserver.ranged', True)
else:
log.warning(
u"Cannot satisfy ranges in Range header: %s for content: %s", header_value, unicode(loc)
)
return HttpResponse(status=416) # Requested Range Not Satisfiable
# If Range header is absent or syntactically invalid return a full content response.
if response is None:
response = HttpResponse(content.stream_data())
response['Content-Length'] = content.length
newrelic.agent.add_custom_parameter('contentserver.content_len', content.length)
newrelic.agent.add_custom_parameter('contentserver.content_type', content.content_type)
# "Accept-Ranges: bytes" tells the user that only "bytes" ranges are allowed
response['Accept-Ranges'] = 'bytes'
response['Content-Type'] = content.content_type
# Set any caching headers, and do any response cleanup needed. Based on how much
# middleware we have in place, there's no easy way to use the built-in Django
# utilities and properly sanitize and modify a response to ensure that it is as
# cacheable as possible, which is why we do it ourselves.
self.set_caching_headers(content, response)
return response
def process_request(self, request):
"""Process the given request"""
asset_path = request.path
if self.is_asset_request(request):
# Make sure we can convert this request into a location.
if AssetLocator.CANONICAL_NAMESPACE in asset_path:
asset_path = asset_path.replace('block/', 'block@', 1)
# If this is a versioned request, pull out the digest and chop off the prefix.
requested_digest = None
if StaticContent.is_versioned_asset_path(asset_path):
requested_digest, asset_path = StaticContent.parse_versioned_asset_path(
asset_path)
# Make sure we have a valid location value for this asset.
try:
loc = StaticContent.get_location_from_path(asset_path)
except (InvalidLocationError, InvalidKeyError):
return HttpResponseBadRequest()
# Attempt to load the asset to make sure it exists, and grab the asset digest
# if we're able to load it.
actual_digest = None
try:
content = self.load_asset_from_location(loc)
actual_digest = getattr(content, "content_digest", None)
except (ItemNotFoundError, NotFoundError):
return HttpResponseNotFound()
# If this was a versioned asset, and the digest doesn't match, redirect
# them to the actual version.
if requested_digest is not None and actual_digest is not None and (
actual_digest != requested_digest):
actual_asset_path = StaticContent.add_version_to_asset_path(
asset_path, actual_digest)
return HttpResponsePermanentRedirect(actual_asset_path)
# Set the basics for this request. Make sure that the course key for this
# asset has a run, which old-style courses do not. Otherwise, this will
# explode when the key is serialized to be sent to NR.
safe_course_key = loc.course_key
if safe_course_key.run is None:
safe_course_key = safe_course_key.replace(run='only')
if newrelic:
newrelic.agent.add_custom_parameter('course_id',
safe_course_key)
newrelic.agent.add_custom_parameter('org', loc.org)
newrelic.agent.add_custom_parameter('contentserver.path',
loc.path)
# Figure out if this is a CDN using us as the origin.
is_from_cdn = StaticContentServer.is_cdn_request(request)
newrelic.agent.add_custom_parameter('contentserver.from_cdn',
is_from_cdn)
# Check if this content is locked or not.
locked = self.is_content_locked(content)
newrelic.agent.add_custom_parameter('contentserver.locked',
locked)
# Check that user has access to the content.
if not self.is_user_authorized(request, content, loc):
return HttpResponseForbidden('Unauthorized')
# Figure out if the client sent us a conditional request, and let them know
# if this asset has changed since then.
last_modified_at_str = content.last_modified_at.strftime(
HTTP_DATE_FORMAT)
if 'HTTP_IF_MODIFIED_SINCE' in request.META:
if_modified_since = request.META['HTTP_IF_MODIFIED_SINCE']
if if_modified_since == last_modified_at_str:
return HttpResponseNotModified()
# *** File streaming within a byte range ***
# If a Range is provided, parse Range attribute of the request
# Add Content-Range in the response if Range is structurally correct
# Request -> Range attribute structure: "Range: bytes=first-[last]"
# Response -> Content-Range attribute structure: "Content-Range: bytes first-last/totalLength"
# http://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.35
response = None
if request.META.get('HTTP_RANGE'):
# If we have a StaticContent, get a StaticContentStream. Can't manipulate the bytes otherwise.
if isinstance(content, StaticContent):
content = AssetManager.find(loc, as_stream=True)
header_value = request.META['HTTP_RANGE']
try:
unit, ranges = parse_range_header(header_value,
content.length)
except ValueError as exception:
# If the header field is syntactically invalid it should be ignored.
log.exception(u"%s in Range header: %s for content: %s",
text_type(exception), header_value,
unicode(loc))
else:
if unit != 'bytes':
# Only accept ranges in bytes
log.warning(
u"Unknown unit in Range header: %s for content: %s",
header_value, text_type(loc))
elif len(ranges) > 1:
# According to Http/1.1 spec content for multiple ranges should be sent as a multipart message.
# http://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.16
# But we send back the full content.
log.warning(
u"More than 1 ranges in Range header: %s for content: %s",
header_value, text_type(loc))
else:
first, last = ranges[0]
if 0 <= first <= last < content.length:
# If the byte range is satisfiable
response = HttpResponse(
content.stream_data_in_range(first, last))
response[
'Content-Range'] = 'bytes {first}-{last}/{length}'.format(
first=first,
last=last,
length=content.length)
response['Content-Length'] = str(last - first + 1)
response.status_code = 206 # Partial Content
if newrelic:
newrelic.agent.add_custom_parameter(
'contentserver.ranged', True)
else:
log.warning(
u"Cannot satisfy ranges in Range header: %s for content: %s",
header_value, text_type(loc))
return HttpResponse(
status=416) # Requested Range Not Satisfiable
# If Range header is absent or syntactically invalid return a full content response.
if response is None:
response = HttpResponse(content.stream_data())
response['Content-Length'] = content.length
if newrelic:
newrelic.agent.add_custom_parameter(
'contentserver.content_len', content.length)
newrelic.agent.add_custom_parameter(
'contentserver.content_type', content.content_type)
# "Accept-Ranges: bytes" tells the user that only "bytes" ranges are allowed
response['Accept-Ranges'] = 'bytes'
response['Content-Type'] = content.content_type
# Set any caching headers, and do any response cleanup needed. Based on how much
# middleware we have in place, there's no easy way to use the built-in Django
# utilities and properly sanitize and modify a response to ensure that it is as
# cacheable as possible, which is why we do it ourselves.
self.set_caching_headers(content, response)
return response
